[Espilet]

Hello everyone,

I'm solving an unsteady conjugate heat transfer in a region which consist of one fluid domain and 3 solid domains. The geometry is like a 4 concentric annular cylinders but a little more complicated. For fluid region the k-e method with total energy equation has been chosen. There are 3 interfaces and the GGI method has been chosen for that. All meshes has been created in ICEM CFD separately and then the model is assembled in CFX. For checking the interfaces connectivity the average temperature in each interface in both sides (side 1 and 2) has been plotted versus simulation time. But there are discrepancies between the temperatures on two sides of each interface. The Max. difference is about 2.2 c, 1.5 c and 0.01 c for interface 1,2 and 3 respectively. I checked the heat imbalance between domains but the error was very small. Also the RMS residuals are small as 1e-5. Also The size of interface meshes are nearly in a same order.Is the problem because of inaccuracy of GGI method in matching the meshes? Is there any idea to reduce this error?Thank you so much.

[Antanas]

Originally Posted by Espilet

Hello everyone,

I'm solving an unsteady conjugate heat transfer in a region which consist of one fluid domain and 3 solid domains. The geometry is like a 4 concentric annular cylinders but a little more complicated. For fluid region the k-e method with total energy equation has been chosen. There are 3 interfaces and the GGI method has been chosen for that. All meshes has been created in ICEM CFD separately and then the model is assembled in CFX. For checking the interfaces connectivity the average temperature in each interface in both sides (side 1 and 2) has been plotted versus simulation time. But there are discrepancies between the temperatures on two sides of each interface. The Max. difference is about 2.2 c, 1.5 c and 0.01 c for interface 1,2 and 3 respectively. I checked the heat imbalance between domains but the error was very small. Also the RMS residuals are small as 1e-5. Also The size of interface meshes are nearly in a same order.Is the problem because of inaccuracy of GGI method in matching the meshes? Is there any idea to reduce this error?Thank you so much.

As I remember GGI interface is recommended when conjugate heat transfer is modelled, because it gives no discontinuity in values of temperature across the interface. Have you tried to compare T at a point on interface rather than average T over whole interface surface?

[Espilet]

Originally Posted by Antanas

As I remember GGI interface is recommended when conjugate heat transfer is modelled, because it gives no discontinuity in values of temperature across the interface. Have you tried to compare T at a point on interface rather than average T over whole interface surface?

Yes, I have. This is the plot of temperature after 7200sec along the line y in a plane normal to z which intersects all domains cross sections. As you see there is not any temperature jump. I also tried on another line parallel to x axis but there is not any jump.
The narrow flat region in the top of graph is related to fluid domain.
I also checked the Heat flux and temperature contour on each sides of interface, but there is difference and the max and min value are different in each sides

[Espilet]

Plot after 400 sec

[Espilet]

Although the error in these times is about 1.5 c and it is natural that we can't see this error in a range of 30 c of temperature in these plots
How can I check the temperature in a point in both sides of an interface?

[Antanas]

Originally Posted by Espilet

Although the error in these times is about 1.5 c and it is natural that we can't see this error in a range of 30 c of temperature in these plots
How can I check the temperature in a point in both sides of an interface?

Actually, I meant comparison of Temperature across interface in contrast with comparison of averages over its surface. So you can plot T distribution on a line that crosses the interface. If there's no discontinuities then it's ok.

[Espilet]

Originally Posted by Antanas

Actually, I meant comparison of Temperature across interface in contrast with comparison of averages over its surface. So you can plot T distribution on a line that crosses the interface. If there's no discontinuities then it's ok.

Thank you so much, dear Antanas.
I understood what exactly you said and these two last plots are temp. distributions along the line that crosses all interfaces. I've also tried with another line prepenrdicular to this and there wasn't any discontinuties.